Re: [OPE-L] critics of statistics

From: Rakesh Bhandari (bhandari@BERKELEY.EDU)
Date: Mon May 01 2006 - 13:08:22 EDT


>--------------------- Original Message --------------------
>Subject: critics of statistics
>From:    "Luis A. Aviles" <laviles@uprm.edu>
>Date:    Sun, April 30, 2006 12:59 am
>To:      glevy@pratt.edu
>-----------------------------------------------------------
>
>Dear Jerry:
>
>Please, forward the message, I would like to see what comes from it.
>
>Let me explain something to you about my interest in the matter.  I am
>currently working in a project in which I present a critique of statistics
>based on the results of the U.S census conducted in Puerto Rico, as it
>relates to racial classification.  My work criticizes the positivist
>assumptions of racial statistics.  I would like to know of people working
>in the field of economics, and other fields, who work on issues related to
>a critique of any aspect of the social production of statistics.  As the
>field of statistics presents itself of being non-ideological, I would like
>to be familiar with critiques of statistics.  Do you know of anyone?
>Thanks,
>Luis A. Aviles
>University of Puerto Rico, Mayaguez


I include some of my own reflections on this 
topic, but my thinking has developed since I wrote
some of the below for listserves about eight 
years ago (see archives of LBO-talk); some of 
this was included in my dissertation, and there 
are no footnotes here, so please do not cite 
below.

  But here are three sources:
Tukufu Zuberi, THICKER THAN BLOOD: An Essay on 
how Racial Statistics Lie Minneapolis:University 
of Minnesota Press, 2001)
The Politics of Large Numbers : A History of 
Statistical Reasoning  by Alain Desrosières,
http://raceandgenomics.ssrc.org/ (a set of 
articles on racial classification in medicine)
several pieces by Michael Root on making kinds, classification and race.

DO NOT CITE BELOW WITHOUT MY PERMISSION


ON THE CONCEPT OF GENERAL INTELLIGENCE
Stephen Jay Gould argued against IQ testing on 
the basis of his critique of the factor analytic 
derivation of g (a unitary quality underlying all 
mental cognitive activity) as the principal 
component which can resolve the greatest amount 
of information from a battery of tests. Gould did 
not really argue against the cultural or class 
and race biases of the test-- that is, IQ tests 
are biased to validate status quo hierarchies as 
a reflection of class- and race-differentiated 
innate capacities (perhaps a more suitable topic 
for what is supposed to be a popular critique of 
the IQ industry) --but the attempt to resolve 
test information into a principal component, 
instead of rotating factor axes from their 
principal component to new positions until 
multiple clusters of intelligence are identified 
and g has disappeared.  That is, Gould argued 
that g is a reification, an artifact of a 
statistical technique, and the attempt to 
simplify test information into a principal 
component is ideologically (or metaphorically) 
driven.
This argument which is as easily misinterpreted 
as a positivist stricture against the postulation 
of unobservable explanatory variables has been 
challenged even by other critics of The Bell 
Curve. As Clark Glymour puts it,
Gould claims that factor analysis produces 
conjectures about the existence of unobserved 
properties solely because the properties, if they 
existed, would explain features of data; in his 
phrasing factor analysis 'reifies' unobserved 
quantities, and he thinks 'reification' is a Big 
Mistake. I wonder whether he thinks atoms and 
molecules and their weights are Big Mistakes as 
well, and if not, why not.

However what drove Gould's argument is not 
positivist skepticism towards the postulation of 
unobservable causal mechanisms. Gould would not 
have argued against the reification of g  had it 
be shown to have been detectable in its effects; 
given however that variance in life outcomes that 
general intelligence, much less inherited general 
intelligence, could be shown to account for was 
negligible even in terms of its supporters' own 
data, Gould simply concluded that the postulation 
of its existence has extra-scientific--nay, 
scientistic--grounds. He argued that  Spearman 
had hoped to pass off the method--factor 
analysis-- by which g can be derived from a 
battery of tests as scientific on the basis of 
unjustified analogy to physics:
Thus caught up in physics envy again, Spearman
described his own 'adventurous step of deserting 
all actually observable phenomena of the mind and 
proceeding instead to invent an underlying 
something which --by analogy with physics--has 
been called mental energy.'. Spearman looked to 
the basic property of g--its influence in varying 
degree, upon mental operations--and tried to 
imagine what physical entity best fitted such 
behavior. What else, he argued, but a form of 
energy pervading the entire brain and activating 
a set of specific 'engines,' each with a definite 
locus. The more energy, the more general 
activation, the more intelligence...If g pervades 
the entire cortex, as a general energy, then the 
s-factors for each test must have more definite 
locations. They must represent specific groups of 
neurons, activating in different ways by the 
energy identified with g. The s-factors, Spearman 
wrote (and not merely in metaphor), are engines 
fueled by a circulating g.

        Once the pseudo scientific nature of the 
postulation of g is understood, one is led to 
anthropologist André Béteille's interesting 
comparison: as Indians are (or were)  obsessed 
with ranking people in terms of something that 
cannot be measured (purity and pollution), 
Americans are just as obsessed to the point of 
having developed a compulsory billion dollar 
industry with something just as immeasurable on a 
single hierarchical scale....general 
intelligence!

Pifalls of Racial Classification

Moreover, it is not clear that it is 
scientifically meaningful to racialize the data 
on variance in IQ at all.  While indeed the 
racialization of data can in many cases allow the 
social scientist to measure the effects of racism 
on social processes, the racialization of data 
can arguably contribute in some cases to the very 
racism that the research is meant to play a part 
in remedying--that is, whatever contribution the 
research makes to the eradication of racism 
cannot be segregated from the more important 
contribution it may make to the reification of 
race without which racism would be impossible. 
This threat is especially acute in the case of 
so-called intelligence.
After all, the national controversy over how to 
account for the putative racial IQ gap, created 
in and through the discursive system of racial 
classification, could itself sanction sufficient 
racism and correlative self-doubt as to render 
the possibility of racial causation plausible 
after all: the discursive system of racial 
classification in this case may create the 
conditions  which render its use reasonable as a 
way of measuring its own effect on society. Race 
would seem then to be at times a textbook example 
of what Ian Hacking refers to as dynamic 
nominalism--the process by which systems of 
classification become entrenched, objective and 
perduring.
Implicitly drawing from J.L Austin's speech act 
theory, Pierre Bourdieu has made the same point 
about family discourse:
In a kind of circle, the native category, having 
become a scientific category for demographers, 
sociologists and essentially social workers, who, 
like official statisticians, are invested with 
the capacity to work on reality, helps to give 
real existence to that category. The 
[race]discourse that ethnomethodologists refer to 
is a powerful, performative discourse, which has 
the means of creating the conditions of its own 
verification and therefore its own reinforcement, 
an institutional discourse which durably 
institutes itself in reality.

In the festschrift to the philosopher Nelson 
Goodman, Paul Starr has also expressed the 
problem well: "Categories accumulate. We do not 
ordinarily think about nor act upon the 
categories of social life; we act and think 
within them." This indeed may be the greatest 
tragedy of race. For in this sordid debate about 
racial gaps in intelligence, reflection seems 
required about exactly why there is social 
interest in it in the first place. Why does one 
not find, as Eugene Genovese smartly asked, 
scholarly publications on the IQ gaps between 
Sicilians, WASP's and Jews:
Well, then, why do they [Herrnstein and Murray] 
lump all blacks together? Where, apart from a few 
inadequate and unhelpful remarks, do we find an 
examination of ethnic differences among blacks 
in, say, performance on IQ tests? And the same 
criticism could be extended to the treamtment of 
whites, not all of whom might respond to other 
comparisons with the equanimity they show for 
comparisons involving blacks. Personally, I am 
pleased to be told that blacks are not as smart 
as Sicilians, but I would not recommend that 
anyone try to tell me that Sicilians are not as 
smart as WASP's or Jews.

One cannot naively assume that a racial gap in IQ 
simply exists. In Morris Cohen and Ernest Nagel's 
succinct formulation, classifying " really 
involves, or is a part of the formation of 
hypotheses as to the nature of things." . That 
is, classification choices are themselves part of 
theory choices made in terms of values like 
explanatory power, empirical adequacy, 
simplicity, and so forth, but in particularly 
contested issues like social inequality, 
political values enter in as well. The 
classification of variance in IQ by race has to 
be understood as a choice: the classification may 
imply that the researcher believes that racism or 
congenital racial inferiority is an important 
causal factor in that variance or that the 
researcher believes that the width of the 
dispersion will become either more or less 
politically tolerable if it is understood in 
racial terms. But the state and the social 
sciences, in their official capacity and due to 
their cognitive authority, establish conventions 
of how to see things  Yehudi Webster for example 
would argue that official racialized statistics 
only reify racial categories and overall 
contribute to the very racism that they are 
trying to eradicate--that is, whatever 
contribution they make to the measurement and 
eradication of racism cannot be segregated from 
(as noted above) the more important contribution 
they make to the reification of race without 
which racism would be impossible. While this 
argument may be unpersuasive in many aspects of 
social life, it has greater plausibility in the 
study of "race"-based differences in 
"intelligence." In short, the national debate 
over the racial IQ gap may be one aspect of the 
racism that coupled with wealth poverty would 
have led to a greater "racial" gap if not for the 
resilience of minority culture. One could of 
course argue that if one did not racialize the 
data there would be no way of confirming the 
hypothesis about the resilient superiority of 
minority culture, but this seems to be a case 
that the racism and self doubt generated by 
official debate about how to explain the gap 
have been simply too destructive to want to keep 
the debate alive.
  To reiterate: this is not to suggest an argument 
against the racial classification of data on 
social problems tout court; it is rather an 
argument for self-reflexive awareness from the 
social scientist about why problems are 
understood in racial terms in the first place. 
One can imagine for example Murray and Herrnstein 
jumping on the fact of a higher propensity to 
domestic homicide among blacks than whites. Yet, 
as reported in the Journal of the American 
Medical Association, the six-fold difference in 
black and white rates of domestic homicide 
disappear when household crowding is used as a 
measure of socioeconomic status.  A racial 
classification of the distribution of this social 
pathology thus leads to a mis-specification of 
the causal forces which partially explain 
pathology. Racial comparative statistics on 
domestic homicide rates do not depict a racial 
reality. Rather, they are the data produced by 
the racial theory. But it is more portentous than 
a simple scientific error in classification. What 
implications will be drawn from the "self-evident 
fact" of the more violent propensities of blacks? 
What race-specific solutions will be proposed? 
This is of course not an idle question. In 1992, 
Dr. Frederick Goodwin, then head of the 
Alcoholism, Drug Abuse, and Mental Health 
Administration (ADAMHA), focused on a "rising 
rate of homicide among black men" and later 
attempted to make sense of the ghetto through 
analogies with violent behavior of monkeys in the 
jungle, reading stereotypes of African-Americans 
back into monkeys, as Pat Shipman wryly put it. 
Goodwin would later be involved with the Violence 
Initiative which proposed to study the putative 
biological and chemical deficiencies of those who 
had committed acts of violence and devise 
appropriate drug therapies. It was in short an 
attempt to cast a social issue as a public health 
problem of a racially identifiable group.   It is 
simply not obvious that variance in social 
outcomes is always well understood in racial 
terms and that the social-scientific attempt to 
do so can be easily segregated from the causal 
forces that produce differences in social 
outcomes by race. But this kind of 
self-reflexiveness is underdeveloped among 
positivist social scientists.
Given the nonsensical nature of any claim about a 
heritable racial IQ gap, if not the existence of 
the "fact" of a racial IQ gap,  it is astonishing 
to discover academic ambivalence about the 
possibility of deep heritable racial differences 
in cognitive potential. In the Black White Test 
Score Gap Christopher Jencks and Meredith 
Phillips report:

Snyderman and Rothman asked a sample of over 1000 
psychologists, sociologists, and educational 
researchers, 'Which of the following  best 
characterizes your opinion of the heritability of 
black white differences in IQ?' Of the 661 
'experts' who returned a questionnaire, 14 
percent declined to answer this particular 
question, 24 percent said the data were 
insufficient, 1 percent thought the gap was 'due 
entirely to genetic variation,' 15 percent 
entirely to environmental variation, and 45 
percent thought it was a 'product of both genetic 
and environmental variation.' It is not clear how 
many of those gave the 'both' response would 
accept our conclusion that genes do not play a 
large role in the black-white gap.

This deeply troubling survey result and 
interpretation seem to suggest that most 
educational psychologists, as well as these 
authors, do not rule out the possibility that 
Africans are somewhat deeply racially different 
than whites in their respective capacities for 
intelligent thought. As Block notes, many critics 
of The Bell Curve were implicitly arguing for a 
low degree of genetic inferiority in blacks.  It 
is interesting that while the psychologist Arthur 
Jensen is referred to five times in this volume, 
Richard Lewontin, the world famous geneticist and 
co-author of Education and Class, is not 
mentioned once. There is thus unsurprisingly no 
nuanced discussion of what heritability even 
means and the problems of conducting such an 
analysis on human, as opposed to on plant and 
breeding animal populations. Yet it seems clear 
to me that there has to be extra scientific 
reasons not to dismiss in unequivocal terms the 
idea of any heritable racial difference in terms 
of the pseudo- entity of (reified) intelligence. 
This would be the only reasonable conclusion to 
reach scientifically--the refusal to even 
consider the hypothesis of any heritable racial 
differences in intelligence in terms of 
insufficient evidence to accept the existence of 
the posited entities of deeply differentiated 
races with concordant variations and the factor 
analytic derived entity of g. That is the 
hypothesis of whether there are genetically 
determined IQ differences between the races seems 
a scientific non starter, no less absurd that 
doing a few regressions to reject the hypothesis 
of whether witchcraft truly brings misfortune. 
Indeed we know from Evans-Pritchard's analysis of 
witchcraft of how internally contradictory and 
explanatorily insufficient belief systems can not 
only maintain their purchase but also remain 
important for the reproduction of an unequal 
society.  The power of race, like witchcraft, 
partially derives from how much it allows one to 
make easy sense of various misfortunes and the 
in-built safeguards it has against explanatorily 
failure. However, it is not yet widely accepted 
in the academy that there is as much madness in 
this whole way of thinking about the heritable 
differences in races as there would be in the 
power of witchcraft. To say that there are no 
genetically inferior races is inadequate because 
it commits one at least to that way of speaking. 
It seems to commit one to the view that even if 
there is no decisive evidence for genetically 
inferior races, there might have been or still 
be. To speak in this way about witches is more 
easily seen be involved in a distortion than if 
one speaks this way about heritable racial 
differences.

Race and Medicine

As has been noted by Vincente Navarro, on one of 
the rare occasions (in 1986) that the government 
did collect statistics on mortality rates for one 
of the most important causes--heart disease--by 
class, it was discovered that there was a higher 
differential in the rate for heart disease 
between blue and white collar workers (2.3. times 
higher for the former) than there was between 
blacks and whites (about 1.3 times higher). 
Drawing the conclusion that "growing class 
mortality differentials, which are ignored by the 
government and the media, are primarily 
responsible for the growing race mortality 
differentials, not the other way around," Navarro 
is arguing, in effect, that reliance of race as a 
variable in health statistics  reifies white 
identity while 'de-emphasizing the class 
structure of American society and (making) it 
appear that the differences are genetic (and 
pan-racial) and therefore not a function of 
social and economic inequality or racism." 
Navarro himself writes:

  The evidence is overwhelming that class is an 
extremely important category for understanding 
the lives and deaths of the US population. Why, 
then, do the government and the media focus on 
race and ignore class? Why is the United States 
the only major developed capitalist country that 
does not collect mortality statistics by 
class....The absence of class analysis and class 
discourse is a victory for the capitalist class, 
which encourages the myth of the middle-class 
society. The capitalist class emphasizes race 
rather than class as a means to keep white 
workers on its side. For instance, mortality 
statistics that show that whites have better 
health indicators than blacks suggest that white 
workers are more similar to their white bosses 
than they are to black workers. Because of 
racism, blacks have higher mortality indicators 
than whites within each class and within each 
occupational category. That is why it is 
important to publish mortality statistics by 
race, standardized by class. But the publication 
of mortality statistics by race alone is not only 
unscientific, it is an ideological statement. It 
assumes--as the federal government does--that 
race is the most important category by which to 
divide our population. This assumption is of 
course wrong: it divides rather than unites 
people who, in fact, have more similarities than 
differences in their ways of living and dying. 
(109)


For all its powerful implications, Navarro's 
attempt to reduce race to "an ideological 
statement" is however ultimately unsatisfactory. 
For one, it does not capture the ways in which 
race is subjectively experienced as a 
self-evident category. Secondly, Navarro does not 
dwell on the complexity of how race can be both 
(merely) a misrepresentation of fundamental 
social contradictions and still a (real) 
precipitant of inequality, both an ideology and a 
material force.  That is, Navarro seems to be 
saying both that membership in the minority group 
gives one a greater likelihood of experiencing a 
certain outcome  (the implicit assumption being 
that "it's really minorities' overrepresentation 
in the working class that explains their 
disproportionate health problems.") and that 
minority status in itself is also in some sense  
causally relevant  to being subjected to some 
specified outcomes. In all fairness to Navarro, 
there is certainly inherently nothing 
contradictory in arguing that race is both an 
index of a more powerful factor that cannot for 
ideological reasons be specified in any other way 
and itself a factor in the eventuation of 
outcomes. To be sure, Navarro is not saying that 
minority status is causally relevant only in so 
far as it gives one a greater statistical 
likelihood of suffering blue collar working 
conditions. Race itself can be the proximate 
cause of outcomes through for example stress from 
racism or awareness of the low status to which 
one is consigned in the eyes of others on account 
of this or that somatic particularity.  Navarro's 
point seems to be that not only is race 
overburdened as an explanation for the health 
problems of minorities, it serves to obscure the 
problems of workers, regardless of race. And he 
indeed can say that without having to deny that 
racial discrimination is active in the social 
world.


This archive was generated by hypermail 2.1.5 : Wed May 31 2006 - 00:00:03 EDT